A correlation method was developed to examine functional interactions between brain regions, by correlating either regional cerebral metabolic rates for glucose or regional cerebral blood flows, as determined by positron emission tomography (PET) in humans. In humans in whom regional cerebral blood flow (rCBF) was measured with PET during a face matching task, correlations between visual brain areas and frontal regions were reduced in patients with mild dementia of the Alzheimer's type (DAT) compared to controls. A systems-level neural network-model, fitted to rCBF PET data, permitted determination of the brain regions and their interactions that were involved in two visual processing tasks, and in a working memory task. A multiple regression/discriminant analysis involving PET regional interdependencies distinguished DAT patients from controls.